METRICS: A MEASUREMENT ENVIRONMENT FOR MULTI-CORE TIME CRITICAL SYSTEMS

With the upcoming shift from single-core to multicore COTS processors for safety critical products such as avionics, railway or space computer subsystems, the safety critical industry is facing a trade-off in terms of performance versus predictability. In multi-core processors, concurrent accesses to shared hardware resources are generating inter-task or inter-application timing interference,  breaking the timing isolation principles required by the standards for such critical software. Several solutions have been proposed in the literature to control or regulate these timing interference, but most of these solutions require to perform some level of profiling, monitoring or dimensioning. As time-critical software is running on top of Real Time Operating Systems (ROTS), classical profiling techniques relying on interrupts, multi-threading, or OS modules are either not available or prohibited for predictability, safety or security reasons. In this paper we present METrICS, a measurement environment for multi-core time-critical systems running on top of the industry-standard PikeOS RTOS. Our framework proposes an accurate real-time runtime and resource usage measurement
while having a negligible impact on timing behaviour, allowing us to fully observe and characterize timing interference. Beyond being able to characterize timing interference, we evaluated METrICS in term of accuracy of the timing and resource usage  easurements, intrusiveness both in term of timing and impact on the legacy code. We also present a portfolio of the kind of measurements  METrICS provides.